I took a minute to read the licensing guide for vSphere 5 and I'm still trying to pull my jaw off the floor. VMware has completely screwed their customers this time. Why?
What I used to be able to do with 2 CPU licenses now takes 4. Incredible.
Today
BL460c G7 with 2 sockets and 192G of memory = 2 vSphere Enterprise Plus licenses
DL585 G7 with 4 sockets and 256G of memory = 4 vSphere Enterprise Plus licenses
Tomorrow
BL460c G7 with 2 sockets and 192G of memory = 4 vSphere Enterprise Plus licenses
BL585 G7 with 4 sockets and 256G of memory = 6 vSphere Enterprise Plus licenses
So it's almost as if VMware is putting a penalty on density and encouraging users to buy hardware with more sockets rather than less.
I get that the vRAM entitlements are for what you use, not necessarily what you have, but who buys memory and doesn't use it?
Forget the hoopla about a VM with 1 TB of memory. Who in their right mind would deploy that using the new license model? It would take 22 licenses to accommodate! You could go out and buy the physical box for way less than that today, from any hardware vendor.
Anyone else completely shocked by this move?
JAndrews wrote:
aroudnev seems to be concerned that because Microsoft sells Windows they are prevented from supporting anything else. Which would be true if all they were was an OS company.
My point is they support the Mac OS pretty well and are one of the (if not the) top software vendor for Macs.
Feel free to point out the XBox, keyboards etc as proof they are really a hardware company, or 365 as proof they are actually a cloud provider.
I think we're trying to make similar points. I agree that there's nothing to prevent Microsoft from making products outside their core expertise. I also wanted to highlight the fact that an operating system is software, so the distinction to be made is probably between being an application company and an OS company.
That said, aroudnev's point is probably more that Microsoft, unlike VMware and Citrix, has a dog in the OS race, so people may not want to trust Microsoft with their non-Windows workloads. Of course, the same argument could be made about Oracle and Red Hat.
Will licensing impact implementation of Host cache?
http://sostech.wordpress.com/2011/08/19/ssd-overcommit-and-vsphere5-licensing/
JAndrews wrote:
Will licensing impact implementation of Host cache?
http://sostech.wordpress.com/2011/08/19/ssd-overcommit-and-vsphere5-licensing/
vRAM entitlement : $39/GB (std) / $54/GB (ent) / $44/GB (ent+)backed by physical RAM (pRAM): $35/GBbacked by RAID1 SSD: $40/GB (SATA) / $60/GB (SAS)SSD swap is still ridiculously slower than RAM speeds (remember, we removed the FSB because it was too slow and put the memory controllers on CPUs now, going through a south bridge for storage devices alone is a bit slower, on top of that overhead of file systems and disk controllers, and much more limited bandwidth).
IMO: If you're relying on host cache you're doing it wrong, it's nice and I guess bouncing off host cache every once in awhile is ok, but your performance is going to kick the bucket as soon as you swap.
Can I just export RAM disks via iSCSI and use those as swap?
Even with a ~.5ms response time it's probably still way too slow, and way to little bandwidth.
Anyone try building a RAM disk under ESXi?
Also: On the subject of vTax and RAM disks, I've run phone systems on ESXi (bah you non-believers! It can be done! 30+ concurrent phonecalls all day, no one even noticed. We even rightsized the box from a quad-core with 4GB of RAM to a 1vCPU box with 1GB that still sat at 3% usage all day and under 700MB RAM, heh.), one of the big issues you do run into is dumping phone recording buffers to disk which creates a lot of disk seeking and usually will top your box at around 50-60 calls with local disks.
You get around this easily by implementing a RAM disk, and dumping the entire file (then compressed) from RAM to your storage medium in one go. There are a lot of other fun tricks you can do with RAM disks too where you're only using a small one to get better peformance... vTax pretty much makes this very cost-ineffective. ![]()
MS pretends that they supports other OS but in reality they do not - they do not support any of their products on other OS systems (implementations are old, ineffective, slow urgly and so on).
aroudnev wrote:
MS pretends that they supports other OS but in reality they do not - they do not support any of their products on other OS systems (implementations are old, ineffective, slow urgly and so on).
SuSE and Redhat are officially supported, but I guess those are old, ineffective and so on. (I agree Linux isn't visually beautiful.
, though I like some of the touches in RHEL 6 builds).
If you mean support fo virtualization - just compare how they implemented wiondows as VM (using para virtualization) and Linux (no difference between rhel and suse in reality) as VM - do they have good virtual tools installed to make system work faster, do they use paravirtualization, use specialized graphical driver and so on?
Remember, Linux require VMware tools to work fast on VMware, so SUPPORT means company supports good analog of XXXtools. We are talking about support as VM on HyperV, of course. When we run Linux on Oracle VM, we use XEN version ofthe kernel (so it has embedded support); when we run Linux @ Vmware, we install VMware tools (and system is not well usable before tools are installed); what about HyperV?
And do they support Linux P2V migration (as Vmware do)? From HW Linux to HyperV? Tool which clone system and add necessary drivers into initrd image?
aroudnev wrote:
If you mean support fo virtualization - just compare how they implemented wiondows as VM (using para virtualization) and Linux (no difference between rhel and suse in reality) as VM - do they have good virtual tools installed to make system work faster, do they use paravirtualization, use specialized graphical driver and so on?
Apparently paravirtualization does not improve performance on the new fancy CPUs, because VMware's dropping it.
So maybe it's not so bad if they don't support Linux paravirt?
Sucks for me, because I run Linux guests using paravirt, and I don't have new fancy CPUs with "EPT / RVI"; and buying new
servers with Nehalems sounds like a really expensive way of getting the performance improvement of paravirt for NTP and the like...
Perhaps if they do support paravirt, their performance will actually exceed VMware's :smileyplain:
"VMware products will cease to provide VMI support. VMI will be phased out of Workstation first; beginning in 2010; followed by vSphere in 2011."
-- http://blogs.vmware.com/guestosguide/2009/09/vmi-retirement.html
It is not true - de facto when you install VMware tools, you use paravirtualization - graphical interface, network, disk access, memory control module - and so on all are replaced by direct drivers in some shape. Old paravirtualization was not implemented by VMware the right way and they dropped it, but de facto VMWare tools provides a very effective paravirtualization and what we run under VMware are not VM systems with HW emulation but are instead para virtual systems.
aroudnev wrote:
It is not true - de facto when you install VMware tools, you use paravirtualization - graphical interface, network, disk access, memory control module - and so on all are replaced by direct drivers in some shape. Old paravirtualization was not implemented by VMware the right way and they dropped it, but de facto VMWare tools provides a very effective paravirtualization and what we run under VMware are not VM systems with HW emulation but are instead para virtual systems.
http://www.vmware.com/files/pdf/VMware_paravirtualization.pdf
"It is important to note for clarity that the VMware tools service and the vmxnet device driver are not CPU paravirtualization solutions. They are minimal,
Hardware Assisted Virtualization
"Hardware vendors are rapidly embracing virtualization and developing new features to simplify virtualization techniques. First generation enhancements include Intel Virtualization Technology (VT-x) and AMD’s AMD-V which both target privileged instructions with a new CPU execution mode feature that allows the VMM to run in a new root mode below ring 0.
...
VMware only takes advantage of these first generation hardware features in limited cases such as for 64-bit guest support on Intel
processors."
http://www.vmware.com/files/pdf/perf-vsphere-monitor_modes.pdf
vSphere5 will only support 64-bit VMMs but still has the same monitor options, allowing for software (binary translation), hardware assist or both - but not VMI any longer so no (limited) true paravirtualization.
THis is correct, but this does not eliminate the fact that VMWare tools do exactly the same as old paravirtualization did - strongly optimize system performance by replacing emulation to the direct drivers.
And VMWare tools are in reality a very important part of VMware.
So the question is, does MS support something similar, which can be installed into SLES9, SLES10, SLES11, RHEL4, RHEL5, Solaris x86, has a very high lever of reliability, has all improved (for VM) drivers, and allows VM systems to run with a very low overhead?
Btw, paravirtualization never worked well in VMware - I experimented with sles10 which had paravirtualization support and it always broke java thread locks so we could use it in the very simple cases only. On the other hand, all old Linux XEN implementation required paravirtualization and worked pretty well, and there are speculations (I just don't know) that HyperV uses some kind of paravirtualization (for Windows VM-s) as well. So, to prove that MS really supports all VM guests, they must do the same - develop and support high quality drivers and modules for these VM-s - and I can;t believe that it is true (am I wrong?)
tomaddox wrote:
Novell also made a key misstep in ignoring TCP/IP for so long. It seems like such a small thing, but the whole industry was moving towards TCP/IP as a networking protocol due to the rise of the Internet, and Novell doggedly stuck to IPX/SPX, which was not able to scale as well. ....
Right. And I see no technological "mistake" quite like that on VMware's part.
I appreciate the comparison, but I think Novell is the wrong choice of companies to
compare VMware's situation with.
I am saying Netscape would be a more apt comparison, and I see VMware as
making a major mistake with vRAM Tax, that brings them more and more in line towards
being a 21st century Netscape.
Remember what Internet Explorer did to Netscape's market share,
by being introduced as a free download when Netscape was a separate purchase?
Nothing immediately.... for the first few versions, Internet Explorer was a joke.
Everyone kept using Netscape, for years.
Remember what happened when IE reached version 3; version 3 was shipped
with the OS, had all the same essential functionality, and Netscape was an extra cost,
with few/no benefits for most users?
Netscape was totally irrelevent within a year....
I think we're all overlooking a very positive outcome from this new licensing model, one which hasn't been explicitly mentioned by VMware but which one can reasonably infer from the basis of the model.
To wit, since the new model is capacity based, it would obviously be unfair and unequitable to require purchasers to pay for unused capacity, so one imagines that, when one does the annual true-up with VMware, VMware will obviously be refunding customers some portion of their licensing costs to compensate for any unallocated vRAM. For example, if I've bought 10 Enterprise Plus licenses, I'm entitled to 960 GB of assigned vRAM. If I assign 1152 GB of vRAM, then I'm on the hook for two additional vSphere licenses. If, on the other hand, I only assign 480 GB of vRAM, I should receive a refund for the unused licenses. After all, since they've moved from instance-based licensing to capacity-based, I should not have to pay for unused capacity.
I, for one, am looking forward to my enormous refund check!
tomaddox wrote:
I, for one, am looking forward to my enormous refund check!
Good luck with that one!
Note tho that since you get to trend over a year (but are required to buy before you hit the avg daily high water mark) you do have the option of pay-as-you go, tho as I've pointed out http://wp.me/p1cl48-83 you'll have to jump through some hoops each time you buy-in.
tomaddox wrote:
I think we're all overlooking a very positive outcome from this new licensing model, one which hasn't been explicitly mentioned by VMware but which one can reasonably infer from the basis of the model.
To wit, since the new model is capacity based, it would obviously be unfair and unequitable to require purchasers to pay for unused capacity, so one imagines that, when one does the annual true-up with VMware, VMware will obviously be refunding customers some portion of their licensing costs to compensate for any unallocated vRAM. For example, if I've bought 10 Enterprise Plus licenses, I'm entitled to 960 GB of assigned vRAM. If I assign 1152 GB of vRAM, then I'm on the hook for two additional vSphere licenses. If, on the other hand, I only assign 480 GB of vRAM, I should receive a refund for the unused licenses. After all, since they've moved from instance-based licensing to capacity-based, I should not have to pay for unused capacity.
I, for one, am looking forward to my enormous refund check!
Ho ho, you think VMWare actually trusts them vRAM model on it's own? It isn't about moving to the vRAM model, it's about a way to charge customers more. vRAM is just an excuse to up the prices of one of the most expensive hypervisors available.
Though good point: You're charging on capacity, so yeah.... where is my refund check?
tomaddox wrote:
After all, since they've moved from instance-based licensing to capacity-based, I should not have to pay for unused capacity.
I, for one, am looking forward to my enormous refund check!
You are incorrect. The new licensing model is a hybrid of per processor and capacity. So no refund for you.
I think you all missed the sarcasm which is pretty thick IMO :smileylaugh:
hmtk1976 wrote:
I think you all missed the sarcasm which is pretty thick IMO :smileylaugh:
While sarcastic, it's honestly a good point, if we're paying for capacity, as VMWare keeps boasting, we should only pay for the capacity we use.
Trying to play the devil's advocate here. Why are users complaining since the license prices per se have not been changed. What if there was not no vRAM tax and the prices had been raised instead ?
I also have a brief survey put together to gather all that has been discussed here in the past couple of weeks. Please provide your inputs. Shouldn't take more than a minute to complete. I will post the results here as soon as I get reasonable number of responses.
Indeed. They shouldn't care if we use a single socket machine or an entire cluster if we pay for vRAM. They want two cakes and eat them both.
